In this research we look into users who had the overdraft product enabled in H1 2022, and understand how they are using the product, by splitting them into usage groups, and understanding how these groups transact while using this product.

These groups are split into 2 usage types:

Number of days in a month using overdraft (days groups)
- not using overdraft
- using overdraft <=10 days in a month
- using overdraft >=11 and <= 27 in a month
- using overdraft >=28 in a month

Percentage of the limit used (usage groups)
- 0% usage (not using overdraft)
- <=20% usage
- 20% to 40% usage
- 40% to 60% usage
- 60% to 80% usage
- 80% to 90% usage
- >90% usage
- in arrears (above 100% usage)

Having these groups in mind, we will be answering the following questions:

How many users do we have in each group?
- How stable are these groups over time?
- How do their Lisbon scores evolve over time?
What are the top 10 outgoing transaction categories for each group?
- Days Groups
- Usage Groups
What are the top 10 incoming transaction categories for each group?
- Days Groups
- Usage Groups

How many users do we have in each group? ¶

Here we can see that 39.4% of users with overdraft enabled in H2 2022 didn't use the product. For the ones that did use overdraft in this timeframe, we can see that our group split led to roughly even groups.

	days_bucket	n_users	perc_users
0	not using od	58332	39.4
1	<=10	27989	18.1
2	>=11 and <= 27	33344	21.1
3	>=28	30594	21.4

As for the usage groups, we can see that 23.4% of all users are using overdraft below up to 20% of their limit. Since we are particularly interested in exploring high overdraft usage, we split the 80%+ group in intervals of 10% instead of 20%, so we can have a more detailed view on the >90% usage users, who correspond to 8.4% of all users.

	usage_buckets	n_users	perc_users
0	0%	58332	39.4
1	<=20%	34960	23.4
2	20% to 40%	12753	8.4
3	40% to 60%	10815	7.1
4	60% to 80%	11265	7.3
5	80% to 90%	7589	5.0
6	> 90%	12594	8.4
7	in arrears	1951	1.1

How stable are these groups over time? ¶

Here we can see that the number of monthly days per group is relatively stable in the 6 months we have looked into, with a small decrease in February, which is of course due to this month being shorter than the remaining ones.

We can also see a similarly stable pattern for the usage buckets.

How do their Lisbon scores evolve over time? ¶

As for rating classes, we can see an interesting shift after April 2022. This corresponds to the launch of the 2.0 version of the service that originates these scores (Lisbon). With this new version, there is a bigger difference between these groups and the >=28 days group has an average very close to 12, the cut-off score for eligibility.

We see a similar polarizing effect after the Lisbon 2.0 launch when it comes to usage buckets. This time around, we can see that on average 80%+ usage users are above 12, meaning once again that on average these high intensity users would no longer be eligible for the product after the Lisbon 2.0 Launch.

What are the top 10 outgoing transaction categories for each group? ¶

There are a few transformations we decided to do in order to make this metric meaningful. First, as for transaction categories, we started from the transaction types ( here you can find the details for each of those types). Since a lot of the big outgoing transaction groups such as Card Presentments (PTs) and Direct Transfers didn't give us much detail, we added more granularity to these:

Added MCC categories for card transactions (PTs);
Added Amherst categories (categories from our internal transaction classifier service) to Direct Transfers and Direct Debits;
Split loan repayment transactions into its own group;
Added fees categorizations from our fees model;
Split spaces transactions into their own group.

Also, in order to exclude outliers where users have only one or two transactions corresponding to very big percentages in their transaction proportions, we're excluding users with less than 6 transactions in the selected 6 month period. This leads to an 11.9% decrease in the user base described above (from 150.3k users to 132.4k users).

Then, in order to understand the impact of each transaction category in user spending, we calculated the percentage of each transaction category for each user (100% being the total outgoing transactions), and then calculated the average for all users in each group. We split these results into 2 dimensions: number of transactions and transaction volume.

Note: For each of the charts below, we only see the top 10 transactions per group. Whenever we see empty transaction categories there, it doesn't mean we don't have transactions there, it just means that they didn't make it to the top 10 for that group.

When it comes to the number of transactions, we can see that:

For all groups, the most frequent transaction category is grocery market card transactions, ranging between 14% and 15.8%.
Users who use the overdraft product for more than 11 days have gambling card transactions, gambling fees, money cash financial and ATM in the top 10 categories, whereas users who don't use overdraft don't have these categories in their top 10 (you can find a list of the top 10 merchants for gambling and money cash financial in our annex )
Users that either don't use overdraft or use it for less than 10 days have outgoing spaces transactions in their top 10, which is not the case for the remaining groups.

As for the transaction volume, we see the following patterns:

One of the main differences we see from the charts above is that grocery market card transactions no longer come up first for most groups, meaning that these are generally high quantity, low volume transactions.
Money cash financial transactions come up leading for all groups that use overdraft for 11 days or more, whereas it doesn't make to the top 10 of the users that don't use this product or use it for less than 10 days.
We also see that overdraft users have ATM and gambling transactions in their top 10, which is not the case for users who don't use overdraft.

When it comes to the number of transactions, we can see some interesting patterns that split users that use none or less than 90% of their overdraft limit, and the ones that use more than 90% are in arrears:

For the ones up until 90% of usage, grocery market card transaction are the leading ones in the top 10
For >90% users, gambling transactions take the first place.
For users in arrears, 5 out of the top 10 categories are fees, the highest category is membership fees with an average of 27.5%.

When looking at the transaction volumes, we see some different patterns:

For the groups that are not using overdraft or using it below 40%, the top 3 categories are rent, spaces and uncategorized Direct Transfers.
For groups that use overdraft above 40% but are not in arrears, the top 3 transaction categories become rent, money cash financial, and ATM transactions.
Overdraft fees go from not being part of the top 10 up to 100% usage, to increasing considerably in arrears, becoming once again one the 2nd biggest category for the arrears group with 33.6% of all outgoing transaction volume in this group - the biggest bucket being Wise Transfers (Foreign Currency Transfers) with 39.9%.

What are the top 10 incoming transaction categories for each group? ¶

Here we apply the same logic as above, but only for incoming transactions (e.g. considering Credit Transfers instead of Direct Transfers and Direct Debits).

When it comes to the number of incoming transactions, we can see 2 main patterns:

The vast majority of transactions come from uncategorized Credit Transfers
Spaces is the 2nd biggest for all groups except for the >=28 group, which has uncategorized income, moneybeam and card transaction refunds ahead of spaces.

As for the transaction volume we can see:

An even bigger share of uncategorized credit transfers, with 85+% for all groups.
Spaces and salary income take the next 2 places in the top 10 in the groups up to 27 days of usage.
For the >=28 user group, moneybeam transactions take the 2nd place.

When looking at number of transactions for usage groups, the following patterns emerge:

Uncategorized credit transfers are the biggest bucket.
Spaces, uncategorized income, and moneybeam transactions are respectively 2nd, 3rd and 4th categories for groups with usage up to 80% - after that, between 80% and 100%, spaces transactions fall to 4th place.
As for the arrears group, we have uncategorized income, salary income, and moneybeam taking respectively the 2nd, 3rd and 4th positions.

And finally, we can see the following for the transaction volumes:

Unsurprisingly by now, uncategorized credit transfers take the lead by far.
Spaces take the 2nd place for usage up to 20%.
Between 20% and 90% usage, salary income transactions take the 2nd place.
And finally, we can see that moneybeam transactions generally increase the higher the usage buckets, taking 2nd place for >90% usage and for users in arrears.

Who are the top 10 merchants behind 'gambling_gaming' and 'money_cash_financial' transactions? ¶

	mcc_category	merchant_name	sum_n_txns	n_users	rn
0	gambling_gaming	WIN2DAY	15484	824	1
1	gambling_gaming	TIPICO	14058	866	2
2	gambling_gaming	LOTTOLAND	9189	181	3
3	gambling_gaming	POKERSTARS	8022	672	4
4	gambling_gaming	Red Rhino Limited	7954	442	5
5	gambling_gaming	BET365	7693	455	6
6	gambling_gaming	BWIN.DE	7641	364	7
7	gambling_gaming	Lotto24	7478	795	8
8	gambling_gaming	ROL*Wunderino	4019	280	9
9	gambling_gaming	PAYPAL *HILLSIDESPO	3516	167	10
10	money_cash_financial	Wise	27058	5460	1
11	money_cash_financial	Binance	17302	3772	2
12	money_cash_financial	CRO	13399	2094	3
13	money_cash_financial	CRO Topup	12808	1021	4
14	money_cash_financial	COINBASE IRELAND	8511	1619	5
15	money_cash_financial	CB PAYMENTS EUR	7368	1696	6
16	money_cash_financial	Taptap Send BE	5749	356	7
17	money_cash_financial	binance.com	5315	1641	8
18	money_cash_financial	PAYPAL *HVV	3679	620	9
19	money_cash_financial	crypto.com	3542	984	10